Evaluation of Automatic Hypocenter Determination in the JMA Unified

Tamaribuchi Earth, Planets and Space (2018) 70:141 https://doi.org/10.1186/s40623-018-0915-4

FULL PAPER Open Access Evaluation of automatic hypocenter determination in the JMA unifed catalog Koji Tamaribuchi*

Abstract The Japan Meteorological Agency (JMA) unifed seismic catalog has been widely used for research and disaster prevention purposes for more than 20 years. Since the introduction in April 2016 of an improved method of automatic hypocenter determinations (PF method), the number of detected earthquakes has almost doubled due to a decrease in the completeness magnitude around the Tohoku region, where seismicity has been very active in the aftermath of the 2011 Tohoku earthquake. Automatically processed hypocenters of small events, accepted without manual modifcation, now make up approximately 70% of new events in the JMA unifed catalog. In this paper, we show that the introduction of automated processing did not systematically bias the quality of the JMA unifed catalog. Approxi- mately 90% of automatically processed hypocenters were less than 1 km from their manually reviewed locations in inland and shallow areas. We also considered the use of automated event characterization in real-time monitoring of earthquake sequences using the example of the April 2016 Kumamoto earthquake sequence, when the PF method could have supplied the catalog with about 70,000 events in real time over the course of 2 months. We show that the PF method is capable of monitoring the migration or expansion of the hypocentral distribution and can support statistical analyses such as variations of the b-value distribution. Further improvements in automatic hypocenter determination will contribute to a better understanding of seismicity as well as rapid risk assessment, especially in cases of swarms and aftershocks. Keywords: JMA unifed seismic catalog, Automatic hypocenter determination, 2016 Kumamoto earthquake, Completeness magnitude

Introduction Research Institute for Earth Science and Disaster Resil- A high-quality earthquake catalog, consistently com- ience (NIED), universities, and other institutes. Te JMA piled over a wide spatiotemporal range, is an important unifed catalog is published and routinely updated on the research tool to analyze the history of seismicity, under- web (JMA 2018) and is widely used for research, promo- stand earthquake mechanisms, and elucidate the under- tion of disaster prevention activities, and other purposes. ground structure of seismogenic regions. Te Japan Te 2011 Mw 9.0 Tohoku earthquake produced an Meteorological Agency (JMA) has recorded earthquake enormous number of aftershocks and induced earth- data in Japan for over 100 years (e.g., Schorlemmer quakes in and around Japan. Because all hypocenters et al. 2018). In October 1997, after the 1995 Kobe earth- required a visual check before being added to the catalog, quake, the JMA started to produce a consistent catalog publication of the catalog was considerably delayed. To of seismic events (hereafter the JMA unifed catalog) cope with this situation, the catalog was limited to events in cooperation with the Ministry of Education, Cul- within the Tohoku region of M ≥ 2 (JMA magnitude) ture, Sports, Science and Technology (MEXT), by using for inland locations and M ≥ 3 for ofshore areas. Nev- seismic waveforms recorded by the JMA, the National ertheless, it took over 2 years to determine over 250,000 hypocenters that occurred during 2011. In addition, the *Correspondence: ktamarib@mri‑jma.go.jp number of monthly events in the JMA unifed catalog Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, doubled after 2000 (Fig. 1) with the deployment of Hi-net Japan observation stations by NIED from 2000 to 2002 (Okada

© The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 2 of 10

the automatically estimated hypocenter is cataloged All events M ≥ 1 when the inspector fnds that it was properly determined. M ≥ 2 Hypocenters that are not properly determined can still be cataloged after a simple review. With this change in procedure, the magnitude limitation in the Tohoku earthquake aftershock area was canceled at the end of March 2016. Enough time has elapsed since the change in the catalog procedure to evaluate whether the new procedure caused a systematic bias in the quality of the JMA unifed catalog. Tis evaluation, reported in this paper, is important to ensure the continued reliability of the JMA unifed catalog for seismic research. In addition, it is important to evaluate the use of initial hypocenter determinations Year (before visual inspection) for real-time monitoring of Fig. 1 Monthly frequency of events in the JMA unifed catalog seismic activity, especially for the abundant aftershocks since 1990. Blue bars indicate the number of earthquakes. The red line indicates the number of stations. Automatic hypocenter of large events. In this paper, we use the M 7.3 Kuma- determinations using the PF method (Tamaribuchi et al. 2016) were moto earthquake sequence in April and May 2016 as a introduced in April 2016 case study.

Methods Te PF method was described by Tamaribuchi et al. et al. 2004). Te number of detected earthquakes will (2016) in Japanese, so for convenience a description in further increase with the deployment of seafoor seis- English is presented in Additional fle 1. Here we will mic observation stations such as those in DONET1 and focus on the conventional group trigger method (GT DONET2, developed by the Japan Agency for Marine- method), previously used for the JMA catalog since its Earth Science and Technology (JAMSTEC) and oper- beginning in October 1997, and describe its problems ated by NIED, and S-net, developed and operated by and their solutions by using the PF method. NIED. To aid prompt processing of large amounts of data Figure 2 presents fow diagrams for the GT and PF from these high-sensitivity seismic networks, the Earth- methods. In the GT method, the ratio of the long-term quake Research Committee (ERC) of the Headquarters average (LTA) to short-term average (STA) amplitude, for Earthquake Research Promotion of Japan established or STA/LTA (e.g., Allen 1978), is calculated for seismic three policies: (1) maintaining the present detection lim- waveforms at each observation station, and when the its of the catalog, (2) cataloging all detected earthquakes, ratio exceeds a threshold value, the observation station is and (3) categorizing events based on a quality control regarded as triggered. When a certain number of stations procedure (ERC 2014). Based on these policies, in April are triggered within a certain time range within a prede- 2016 the JMA adopted the Phase combination Forward fned group of stations (for example, four or more sta- search method (PF method; Tamaribuchi et al. 2016) for tions are triggered within 120 s), it is defned as an event automatic hypocenter determination to streamline pro- detection. At this point, the trigger time at each station is duction of the JMA unifed catalog. Te PF method can taken as the P-phase arrival time and a hypocenter is cal- detect simultaneously occurring earthquakes by using a culated from the P-phases (frst-round hypocenter calcu- Bayesian estimation approach. Details of this method are lation). On the basis of the theoretical travel times from described in “Methods” section and Additional fle 1. this initial hypocenter estimation, P-wave and S-wave Automatically processed hypocenters are added to arrivals are automatically picked again and a revised the JMA unifed catalog after quality control by visual hypocenter is calculated from the P- and S-phases (sec- inspection and the assignment of a status fag to each ond-round hypocenter calculation). Phase-picking is event that characterizes its magnitude and accuracy. Te repeated, based on travel times from successively revised procedure for large events did not change in April 2016; hypocenters, until the hypocenters converge to a stable that is, all events greater than the threshold magnitude location. An advantage of the GT method is that it is (such as Mth 1.7 in shallow inland areas and increasing = possible to exclude local noise that afects a single sta- to Mth 3.5 with distance from land; see also JMA 2018 = tion. Also, even when the hypocenter does not converge, in detail) are visually inspected before inclusion in the an event detection can alert the operators to perform a catalog. For events smaller than the threshold magnitude, visual inspection and correctly determine the hypocenter. Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 3 of 10

Because the GT method assumes that noise rarely a GT method b PF method occurs at the same time as earthquakes, it can ignore isolated local noise. However, when an earthquake and Seismic Seismic noise coincide, or when two or more earthquakes occur waveform waveform simultaneously, the method leads to incorrect locations in the frst-round hypocenter calculation. Also, when many earthquakes occur, such as during aftershocks and Trigger by STA/LTA Extract feature swarms, the LTA value is high, so the detection capabil- (P-wave assumed) (read P- or S-phase) ity of the method is degraded. Consequently, after large earthquakes, the GT method failed to detect some events and operators were required to check continuous seismic Group trigger? P ≥ 3? waveforms visually. Te PF method has two characteristic functions: (1) a feature extraction step that chooses candidate P- or Round 1 hypocenter Search optimal calculation (use S-phase arrival times without reference to STA/LTA, phase combination triggered station) and (2) a hypocenter determination step that chooses the optimal combination of phases from the candidate arrival Round 2 hypocenter Round 2 times and the maximum amplitudes recorded near those calculation (use both hypocenter times. Details of the method are described in Additional P and S records) calculation fle 1. Because the PF method is relatively defcient in detect- ing slow earthquakes and in determining hypocenters in Quality check? Quality check? island areas such as the Okinawa region, both the GT and PF methods are used today to detect events and report them to the operator. Automatic Automatic hypocenter hypocenter Figure 3 shows the detailed workfow of event iden- tifcations, and the fags assigned to each event by the operator are summarized in Table 1. For further details, Fig. 2 Flowchart of automatic hypocenter estimation processes: a GT method and b PF method refer to the online publication of the JMA unifed catalog (JMA 2018). For automatically detected events that are larger than the threshold magnitude, the operator inspects the phases and does a full conventional

a Conventional flow (through March 2016)

Seismic Automatic process JMA unified Event list Full review waveform by GT method catalog

b Current flow (after April 2016) Confirm Flag A,a Auto-hypo automated Auto-hypo: good (M < Mth) hypocenter Automatic process Auto-hypo: bad by GT method Seismic Auto- Flag k,s JMA unified Event list Simple review waveform hypo M < Mth catalog Automatic process undefined M≥ Mth by PF method Auto-hypo Flag K,S (M ≥ Mth) Full review

Fig. 3 Flowchart of the production of the JMA unifed catalog: a conventional fow (before April 2016) and b current fow (after April 2016). “Auto-hypo” means the automatically processed hypocenter. The fow from “Event list” to “Simple review” indicates the GT event without automatically processed hypocenter Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 4 of 10

Table 1 Hypocenter fags year-long period before March 2011 with the 21-month Flag Determination method Accuracy period after April 2016 in an efort to avoid the infu- ence of the March 2011 Tohoku earthquake. It shows K Fully reviewed High that Mc was slightly higher than it was before the earth- S Fully reviewed Low quake in and near the rupture area, but slightly lower k Simply reviewed High elsewhere. Tat was because the seismicity was still s Simply reviewed Low high in and near the rupture area (Additional fle 1: Fig- A Automatically processed High ure S2). a Automatically processed Low An analysis of automatically processed hypocenters in the JMA unifed catalog from April 2016 to Decem- ber 2017 (Fig. 5) shows that 63.7% of the events added to the JMA unifed catalog had automatically pro- hypocenter calculation (“fully reviewed hypocenter”) cessed hypocenters (58.5% for fag “A”, 5.2% for fag and assigns it a quality fag of “K” or “S” for high or low “a”). Among smaller events (M < Mth), 73.2% had auto- accuracy, respectively. For events that are smaller than matically processed hypocenters without modifcation the threshold magnitude and for which the automati- (67.3% for fag “A”, 5.9% for fag “a”), signifying a great cally processed phases appear valid, the automatically improvement in the efciency of the JMA unifed cata- processed hypocenter is cataloged without modifcation log procedure. Because all hypocenters with M Mth (“automatically processed hypocenter”) and the opera- ≥ must be fully reviewed, all larger events were fagged tor assigns it a fag of “A” or “a” for high or low accuracy, “K” or “S”. respectively. If the determination is not good, the opera- As a further check, we compared events between April tor inspects and corrects the appropriate phases and 2016 and December 2017 that were fagged “K” (fully then calculates the hypocenter (“simply reviewed hypo- reviewed) in the JMA unifed catalog and the hypocenters center”). Small events are then assigned a fag of “k” or derived by the PF method before their visual inspection. “s” for high or low accuracy, respectively, and large events For this comparison, we chose hypocenter pairs with are fully reviewed as previously described. In this way, all epicenters 50 km apart and within 60 s of each other. hypocenters are classifed on the basis of the detection ≤ Among inland and shallow earthquakes (depth 30 km), method and their accuracy, while the capability of con- ≤ which were surrounded by a seismic observation network ventional hypocenter detection is preserved. giving good azimuthal coverage, approximately 90% of Results and evaluation automatically processed events were within 1 km of the fully reviewed hypocenters (Fig. 6 and Table 2). Among Te monthly number of detected hypocenters nearly ofshore or deep earthquakes, approximately 90% of auto- doubled after the PF method was introduced (Fig. 1). matically processed events were within 5 km of the fully Although aftershocks of the 2016 Kumamoto earth- reviewed hypocenters. Tese results show that the cur- quake immediately after the change took place in April rent seismic observation network, installed at an interval 2016 boosted the number of detected hypocenters, the of about 20 km, can obtain automatically processed hypo- increase persisted through the end of our analytical centers with acceptable reliability. Likewise, there was lit- period in December 2017. tle or no bias in the automatic magnitude determinations To confrm that the new method improves the detec- in almost all areas (Fig. 6b). Te automatic hypocenters tion of seismic events, we performed a detailed analy- near Taiwan had relatively high magnitudes compared sis of the JMA unifed catalog in terms of completeness with the JMA catalog because data delays prevented the magnitude M , the event magnitude above which all c use of IRIS stations in the automatic process. Nearly all events are reliably detected, using the entire-magni- hypocenters in the Taiwan region therefore required a tude-range (EMR) method proposed by Woessner and full review. In the same way, we also compared fagged Wiemer (2005). In a Monte Carlo approximation of “k” (simply reviewed) and the hypocenters derived by the the bootstrap method, we calculated the mean value PF method (Additional fle 1: Table S1). Standard devia- of M from 200 trials, each of which resampled the c tions of hypocentral location and magnitude are larger catalog by picking at least 50 hypocenters at depths than those of fully reviewed hypocenters (Table 2). Also, 50 km, in 1.0° 1.0° grid cells overlapping by 0.5°. ≤ × magnitudes of automatic processing tend to be 0.1 unit Figure 4a shows maps of M values before and after c greater than those of simply reviewed hypocenters on the procedure changed in April 2016. M became c average. We need not elaborate on this point because the lower in a few places, particularly in coastal areas near simply review is required only for inaccurate automati- the Tohoku earthquake rupture zone, but remained cally processed hypocenters. largely unchanged elsewhere. Figure 4b compares the Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 5 of 10

a Comparison of periods before and after the 2016 earthquake

(A) Apr. 2015 (B) Apr. 2016 (C) B –A to Mar. 2016 to Dec. 2017

b Comparison of periods before and after the 2011 Tohoku earthquake

(A) Mar. 2010 (B) Apr. 2016 (C) B –A to Feb. 2011 to Dec. 2017

Fig. 4 Distribution of completeness magnitude (Mc); depth range 0–50 km. a Comparison of the periods April 2015 to March 2016 and April 2016 to December 2017, before and after the introduction of the automatic process. b Comparison of March 2010 to February 2011 and April 2016 to December 2017 (before and after the 2011 Tohoku earthquake)

Application to the 2016 Kumamoto earthquake All It is instructive to examine the efect of the revised process in the case of the 2016 Kumamoto earthquake, M < Mth which occurred immediately after the PF method was M ≥ Mth incorporated. Figure 7a shows the epicentral distribution of the automatically processed hypocenters. Te dataset 0% 20% 40% 60%80% 100% spans the period from April 14 to May 31, 2016. To sin- K S k s A a gle out the efect of the automatic process in real time, Fig. 5 Proportion of fags in the JMA unifed catalog since April 2016. we used the provisional JMA unifed catalog of June 6, All, full range of magnitudes; M < Mth, small events not requiring full 2016, as a reference. Because events in the JMA unifed inspection; M Mth, relatively large events requiring full inspection ≥ catalog must undergo a visual inspection, the number Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 6 of 10

a Location residual b Magnitude residual April 2016 to December 2017 April 2016 to December 2017

Fig. 6 Distribution of a location and b magnitude residuals between fully reviewed hypocenters in the JMA unifed catalog (fag K, depth range 0–50 km) and their associated automatically processed hypocenters (before visual inspection) from April 2016 to December 2017

Table 2 Residuals between fully reviewed events (fag “K”) in the JMA catalog and their corresponding automatically processed hypocenters by the PF method: a averages (ave.) and standard deviations (σ), b proportions of automatically processed hypocenters within 1 and 5 km and 0.1 and 0.5 units of the fully reviewed hypocentral location and magnitude in the JMA unifed catalog, respectively Latitude (km) Longitude (km) Depth (km) Magnitude ave. σ ave. σ ave. σ ave. σ a All events 0.21 4.31 0.10 4.03 0.13 8.57 0.01 0.15 − Shallow inland 0.17 3.29 0.25 3.25 0.09 4.16 0.02 0.15 − − − Deep/ofshore 0.24 4.80 0.30 4.40 0.25 10.28 0.01 0.14 − Latitude Longitude Depth Magnitude < 1 km (%) < 5 km (%) < 1 km (%) < 5 km (%) < 1 km (%) < 5 km (%) < 0.1 unit (%) < 0.5 unit (%) b All events 72.6 93.0 67.7 91.9 52.6 85.7 84.5 99.1 Shallow inland 88.8 95.9 88.9 96.3 67.2 94.4 85.4 98.9 Deep/ofshore 63.3 91.4 55.5 89.4 44.3 80.8 84.0 99.2 of events since 14 April in the provisional catalog as of 6 GT method and an order of magnitude more events than June was 5084, whereas the automatic process detected human reviewers could add to the catalog in the 7 weeks 69,814 events in that time, of which 54,843 events had after the earthquake series began. We stressed that PF time errors within 0.2 s and horizontal errors within 0.5′. method can determine hypocenters within 3 min from During that same period, the GT method detected 18,744 origin time on average. events of which 10,139 had time errors within 0.2 s and Figure 7b shows the spatiotemporal distribution of horizontal errors within 0.5′. In sum, the PF method events, with cataloged events shown in red and auto- determined about fve times as many epicenters as the matically determined events in gray or black. Te Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 7 of 10

a b 16 April 2016 A 01:25 12 km M 7.3

14 Apr 2016 21:26 11 km M 6.5

15 Apr 2016 00:03 7 km M 6.4 B B

c d 100000 JMA catalog 10000 Auto-hypo

1000 of events 100

Number 10

1 012345678 Magnitude Fig. 7 Hypocentral distribution and frequency–magnitude distribution of the 2016 Kumamoto earthquake sequence for events shallower than 50 km (21:00 April 14–24:00 May 31, 2018; depth range 0–50 km). a Epicentral distribution (21:00 April 14 to 24:00 May 31, 2018). b Spatiotemporal distribution (21:00 April 14–24:00 April 17, 2018) projected onto line A–B in a. c Magnitude–time diagram (21:00 April 14–24:00 May 31, 2018). Red dots along the bottom indicate events whose magnitude undefned. Red, black, and gray dots indicate events in the JMA unifed catalog (as of June 6, 2016), high-quality automatically processed hypocenters (time error within 0.2 s and horizontal error within 0.5′), and all automatically processed hypocenters, respectively. d Frequency–magnitude distribution of events related to the 2016 Kumamoto earthquake. Solid symbols indicate events in the JMA unifed catalog, and open symbols indicate automatically processed hypocenters. Circles and triangles indicate cumulative and incremental frequencies, respectively

length of the area of seismicity, as bounded by the auto- PF method makes it possible to follow changes in the matically processed hypocenters, expanded from 20 km hypocentral distribution in real time. immediately after the M 6.5 foreshock at 21:26 14 April Figure 7d graphs the frequency versus magnitude dis- to about 35 km at 01:25 16 April, just before the M 7.3 tribution of the automatically processed hypocenters mainshock. Tis trend was also reported by Kato et al. and the events in the provisional JMA unifed catalog of (2016), who used the matched flter technique (Gib- June 6, 2016. Te two frequency curves are almost identi- bons and Ringdal 2006). Tis result shows that the cal from M 4 to M 2, but the curve for the JMA unifed Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 8 of 10

catalog departs from the Gutenberg–Richter law below the automatically processed catalog. We calculated the M 2 because the catalog did not list small events, whereas b-values at points separated by 0.05º, using 100 or more automatic processing can detect events as small as M 0. hypocenters within 5 km. Te abundance of small events obtained in real time Before the 2016 Kumamoto earthquake sequence, permits a detailed analysis of seismic activity. Here, we the b-value along the Futagawa–Hinagu fault zone investigated the b-value distribution of the frequency– was lower than in other areas (Fig. 8a). Whereas the magnitude distribution, using the maximum likelihood b-value along the fault zone further decreased after method (Aki 1965; Utsu 1965) and estimating Mc using the M 6.5 foreshock on 14 April, it increased after the the EMR method (Woessner and Wiemer 2005). To M 7.3 mainshock on 16 April. Between 18 April and estimate the uncertainty of the b-value, we adopted a 16 May, the b-value along the Futagawa fault zone, the Monte Carlo approximation of the bootstrap method, site of the mainshock rupture (Asano and Iwata 2016; calculating the mean and variance of the b-values in JMA 2017), recovered to its pre-earthquake value. It 200 trials using resamples (with replacement) from remained low in the southern part of the Hinagu fault

Fig. 8 Distribution of b-values before and during the 2016 Kumamoto earthquake sequence (depth range 0–50 km). a January 1, 2006, to April 13, 2016 (before the earthquake), from the JMA unifed catalog. Automatically processed hypocenters (PF method) are shown for b April 14–15, 2016 (including M 6.5 and M 6.4 foreshocks), c April 16–17, 2016 (including M 7.3 mainshock), and d April 18 to May 16, 2016 Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 9 of 10

zone (Fig. 8d), which suggests that stress is still high Abbreviations ETAS: epidemic-type aftershock sequence; GT: group trigger; JMA: Japan Mete- there (Scholz 1968). orological Agency; EMR: entire-magnitude-range; ERC: Earthquake Research Committee; M: magnitude; Mc: completeness magnitude; MEXT: Ministry of Education, Culture, Sports, Science and Technology; NIED: National Research Discussion and conclusion Institute for Earth Science and Disaster Resilience; Mth: threshold magnitude; We have shown that automatic hypocenter determina- PF: Phase combination Forward search. tion by the PF method resulted in dramatic gains in pro- Authors’ contributions duction efciency of the JMA unifed catalog and made KT performed the analysis and wrote the manuscript. The author read and possible a new means of real-time monitoring of seismic approved the fnal manuscript. activity. Tis capability promises to contribute to future forecasts of earthquake probability, as well as rapid Acknowledgements assessment of the extent of aftershock areas and changes We thank Akio Katsumata for internal reviews and helpful comments. The in the extent of seismic activity. Combined with quality earthquake catalog is produced by the JMA in cooperation with MEXT. Values of Mc were estimated with an R program coded by Mignan and Woessner inspection by human experts, this method yields event (2012). We used the generic mapping tools (GMT) package (Wessel et al. 2013) determinations as accurate as those from conventional and hypdsp (Yokoyama 1997) to draw the fgures. We are also grateful to Chung-Han Chan, an anonymous reviewer, and Editor Takuto Maeda for their methods. constructive comments. Tese automatically processed hypocenters, even before quality inspection, aid analyses of aftershock Competing interests activities immediately after large earthquakes. In fact, The author declares no competing interests. the JMA’s frst press release, issued 2 h after the M 6.5 Availability of data and materials foreshock at 23:30 14 April (JMA 2016), relied on unin- The data used in this article are available at the Data Management Center of NIED (http://www.hinet.bosai.go.jp/?LANG en) and JMA (http://www.data. spected hypocenters for the explanation of seismic activ- jma.go.jp/svd/eqev/data/bulletin/index_e.html= ). ity and its relation to the causative fault zone. Tis automatic process may also lend itself to forecasts Consent for publication of seismic activity. For example, Omi et al. (2013, 2015, Not applicable. 2016) demonstrated real-time forecasts of aftershock Ethics approval and consent to participate probability by using Bayesian estimates based on the Not applicable.

Omori–Utsu formula and the epidemic-type aftershock Funding sequence (ETAS) model (Ogata 1988). Future research None. should lead to more rapid risk assessment of aftershocks and foreshocks. Publisher’s Note Tere are obvious avenues for further improvement of Springer Nature remains neutral with regard to jurisdictional claims in pub- this automatic process. For example, hypocenter accuracy lished maps and institutional afliations. can beneft from high-precision hypocenter determina- Received: 2 July 2018 Accepted: 23 August 2018 tion techniques such as the double-diference method (Waldhauser and Ellsworth 2000; Yano et al. 2017) and utilization of three-dimensional velocity structures (Kat- sumata 2015), which are growing in importance with the References enhancement of ocean bottom seismometer networks Aki K (1965) Maximum likelihood estimate of b in the formula log (N) a bM and its confdence limits. Bull Earthq Res Inst Tokyo Univ 43:237–239= − (S-net, DONET). In addition, low-frequency earthquakes Allen RV (1978) Automatic earthquake recognition and timing from single are not determined by this method. Alternative auto- traces. Bull Seismol Soc Am 68:1521–1532 matic hypocenter determination processes such as the Asano K, Iwata T (2016) Source rupture processes of the foreshock and mainshock in the 2016 Kumamoto earthquake sequence estimated from the matched flter technique (Gibbons and Ringdal 2006; kinematic waveform inversion of strong motion data. Earth Planets Space Shelly et al. 2007; Moriwaki 2017) will be important for 68:147. https://doi.org/10.1186/s40623-016-0519-9 those events. Earthquake Research Committee (2014) Report: the improvement of the processing method for the high sensitivity seismic observation data (in Japanese). https://www.jishin.go.jp/main/kokando/kokando_honpe n.pdf. Accessed 2 May 2018 Gibbons SJ, Ringdal F (2006) The detection of low magnitude seismic events Additional fle using array-based waveform correlation. Geophys J Int 165:149–166. https://doi.org/10.1111/j.1365-246X.2006.02865.x Additional fle 1: Text S1. Description of PF method. Figure S1. Flow- Japan Meteorological Agency (2016) Press release: the 21:26, April 14, 2016 chart of hypocenter determination process. Figure S2 Temporal variation earthquake (in Japanese). https://www.jma.go.jp/jma/press/1604/14a/ of completeness magnitude (Mc). Table S1. Residuals between simply kaisetsu201604142330.pdf. Accessed 2 May 2018 reviewed events (fag “k”) in the JMA catalog and their corresponding Japan Meteorological Agency (2017) The source process analysis of the April automatically processed hypocenters by the PF method. 16, 2016 Kumamoto Earthquake using strong motion (in Japanese). Tamaribuchi Earth, Planets and Space (2018) 70:141 Page 10 of 10

https://www.data.jma.go.jp/svd/eqev/data/sourceprocess/event/20160 Scholz CH (1968) The frequency–magnitude relation of microfracturing in rock 41601250547near.pdf. Accessed 2 May 2018 and its relation to earthquake. Bull Seismol Soc Am 58(1):399–415 Japan Meteorological Agency (2018) The seismological bulletin of Japan. https Schorlemmer D, Hirata N, Ishigaki Y, Doi K, Nanjo KZ, Tsuruoka H, Beutin T, ://www.data.jma.go.jp/svd/eqev/data/bulletin/index_e.html. Accessed 2 Euchner F (2018) Earthquake detection probabilities in Japan. Bull Seis- May 2018 mol Soc Am 108(2):702–717. https://doi.org/10.1785/0120170110 Kato A, Fukuda JI, Nakagawa S, Obara K (2016) Foreshock migration preced- Shelly DR, Beroza GC, Ide S (2007) Non-volcanic tremor and low-frequency ing the 2016 Mw 7.0 Kumamoto earthquake, Japan. Geophys Res Lett earthquake swarms. Nature 446:305–307. https://doi.org/10.1038/natur 43(17):8945–8953. https://doi.org/10.1002/2016GL070079 e05666 Katsumata A (2015) Fast hypocenter determination in an inhomogeneous Tamaribuchi K, Moriwaki K, Ueno H, Tsukada S (2016) Automatic hypocenter velocity structure using a 3D travel-time table. Bull Seismol Soc Am determination for the Seismological Bulletin of Japan using Bayesian 106:3203–3208. https://doi.org/10.1785/0120150122 estimation. Quart J Seis 79:1–13 (in Japanese with English abstract) Mignan A, Woessner J (2012) Estimating the magnitude of completeness for Utsu T (1965) A method for determining the value of b in a formula log earthquake catalogs. Community Online Resour Stat Seism Anal. https:// n a bM showing the magnitude–frequency relation for earthquakes. doi.org/10.5078/corssa-00180805 Geophys= − Bull Hokkaido Univ 13:99–103 (in Japanese) Moriwaki K (2017) Automatic detection of low-frequency earthquakes in Waldhauser F, Ellsworth WL (2000) A double-diference earthquake loca- southwest Japan using Matched-Filter technique. Quart J Seis 81:3 (in tion algorithm: method and application to the Northern Hayward Japanese with English abstract) Fault, California. Bull Seismol Soc Am 90(6):1353–1368. https://doi. Ogata Y (1988) Statistical models for earthquake occurrences and residual org/10.1785/0120000006 analysis for point processes. J Am Stat Assoc 83(401):9–27 Wessel P, Smith WH, Scharroo R, Luis J, Wobbe F (2013) Generic mapping tools: Okada Y, Kasahara K, Hori S, Obara K, Sekiguchi S, Fujiwara H, Yamamoto A improved version released. EOS Trans Am Geophys Union 94(45):409– (2004) Recent progress of seismic observation networks in Japan—Hi- 410. https://doi.org/10.1002/2013EO450001 net, F-net, K-NET and KiK-net—. Earth Planets Space 56:xv–xxviii. https:// Woessner J, Wiemer S (2005) Assessing the quality of earthquake catalogues: doi.org/10.1186/bf03353076 estimating the magnitude of completeness and its uncertainty. Bull Omi T, Ogata Y, Hirata Y, Aihara K (2013) Forecasting large aftershocks within Seismol Soc Am 95(2):684–698. https://doi.org/10.1785/0120040007 one day after the main shock. Sci Rep 3:2218. https://doi.org/10.1038/ Yano TE, Takeda T, Matsubara M, Shiomi K (2017) Japan unifed high-resolution srep02218 relocated catalog for earthquakes (JUICE): crustal seismicity beneath the Omi T, Ogata Y, Hirata Y, Aihara K (2015) Intermediate-term forecasting of Japanese Islands. Tectonophysics 702:19–28. https://doi.org/10.1016/j. aftershocks from an early aftershock sequence: Bayesian and ensemble tecto.2017.02.017 forecasting approaches. J Geophys Res Solid Earth 120:2561–2578. https Yokoyama H (1997) Seismic activity analysis program using X window system. ://doi.org/10.1002/2014JB011456 Quart J Seis 60:37–51 (in Japanese) Omi T, Ogata Y, Shiomi K, Enescu B, Sawazaki K, Aihara K (2016) Automatic aftershock forecasting: a test using real-time seismicity data in Japan. Bull Seismol Soc Am 106(6):2450–2458. https://doi.org/10.1785/0120160100